Revert to old analytical likelihood for ddm

[AlexanderFengler]

Somewhere in the process of trying to improve this likelihood we introduced a bug which is still relatively mysterious, but need to revert to not have broken likelihood in main.